为解决无人机在复杂气象条件下易受结冰影响而威胁其飞行安全的问题，提出了一种考虑结冰风险的无人机航迹规划方法。首先，构建基于中尺度WRF （Weather Research and Forecasting）模式的结冰气象预测模型，通过最佳参数化方案组合的结冰气象模拟获得模拟时段内海南乐东地区的温度、压力、液态水含量（LWC）空间分布及时序变化。其次，构建基于代理模型的水滴收集质量快速预测方法。在获取美国联邦航空条例（FAR）25部附录C中连续最大结冰条件下40个采样点处水滴收集质量分布的基础上，利用本征正交分解（POD）降阶模型和Kriging插值算法，建立温度、压力、LWC、平均有效水滴直径（MVD）等结冰气象参数与水滴收集质量之间的代理模型，可快速预测出目标区域内水滴收集质量的空间分布与时序变化。最后，根据飞机结冰强度划分等级，以不同结冰强度下水滴收集质量阈值为结冰安全约束，利用基于粒子群优化（PSO）的结冰容限航迹规划方法进行考虑结冰风险的无人机飞行策略研究。研究结果表明：利用WRF模式可获得温度、压力、LWC等结冰气象参数，预测值与观测值匹配良好；基于POD降阶模型和Kriging插值算法，构建的气象参数与水滴收集质量间代理模型可快速准确地获取目标区域内水滴收集质量的空间分布与时序变化；基于PSO的结冰容限航迹规划方法可在不同结冰安全约束条件下，规划出无人机最优航迹。

UAVs are vulnerable to icing under complex meteorological conditions， thus threatening flight safety. To solve this problem， this paper proposes a UAV trajectory planning method considering the icing risk. Firstly， an icing meteorological prediction model based on the mesoscale Weather Research and Forecasting （WRF） model is constructed. By icing meteorological prediction with the best combination of parameterization schemes， the spatial distribution and temporal evolution of temperature， pressure and Liquid Water Content （LWC） in the Ledong area of Hainan from May to July 2021 are obtained. Secondly， a rapid prediction method for water droplet collection based on the surrogate model is established. The Optimal Latin Hypercube Sampling （OLHS） method is employed to sample the continuous maximum icing conditions in Appendix C of Federal Aviation Regulations （FAR） Part 25， and the droplet impact characteristics are numerically calculated for 40 sampling points to obtain the distribution of water droplet collection at each sampling point. Based on the Proper Orthogonal Decomposition （POD） reduced-order model and the Kriging interpolation method， a surrogate model between the water droplet collection and meteorological parameters， such as temperature， pressure， LWC and droplets Median Volumetric Diameter （MVD）， is established.On the basis of the established surrogate model， the spatial distribution and temporal evolution of water droplet collection in the target area are obtained. Finally， taking the threshold of water droplet collection at various icing intensity as the icing safety constraint， we use the Particle Swarm Optimization （PSO）-based icing tolerance trajectory planning method to optimize the flight strategy of the UAV considering the icing risk to overcome the defects of the existing icing prediction algorithms lack of icing risk quantification. The results show that the icing meteorological parameters predicted by the WRF model， such as temperature， pressure， and LWC， match well with the observations. Based on the POD model and Kriging method， the constructed surrogate model between meteorological parameters and water droplet collection can quickly and accurately predict the spatial distribution and temporal evolution of water droplet collection in the target area. The PSO-based icing tolerance trajectory planning method is competent to plan the optimal trajectory of the UAV under different icing safety constraints.